An intrepid reader from Germany (we're big overseas) writes in to us to note a significant change in the BLS's documentation of the "Birth-Death model". This is a computation the BLS makes to its CES employment statistic to account for supposed business births and deaths which wouldn't be expected to show up in the job creation count. However, the contribution of this factor since 2001 has increased, at times accounting for 40-80% of all jobs tallied as created (discussions can be found here and here).

The BLS used to admit that this factor would tend to be way off during economic trend changes. But as source points out:

Accidentially I noticed yesterday, that the most interesting part of the information concerning the model has been cut out:

Please compare the current content of the site (last updated on Feb 1st 08) to the attached printout I made in Dec 07.

Missing is a part of the text containing sentences like "The most significant drawback to this or any model-based approach is that time series modelling [...] is likely to have some difficulty producing reliable estimates at economic turning points..."

We used to consider the BLS's disclaimers refreshingly honest -- for anyone who actually bothered to read them and go beyond the superficial "headline" employment statistics. For example, these guys even admitted they implemented a "substitution effect" computation -- a favorable downward-smoothing -- into the CPI, even though actual studies looking for the effect were inconclusive. But hey, maybe they've erased that disclosure too. That'll solve the inconsistency.

Apparently, too many people are finding out that the employment statistics are extremely suspect, especially when the economy's trajectory changes. There has been a lot of coverage of this in the blogosphere over the last few years. So right now the BLS must realize that the last thing it needs is to be admitting that its employment numbers become almost purely fictional when entering an economic downturn. Given how important a topic the economy has suddenly become, one might argue this is the most critical time to have accurate numbers! (A point lost on academic economists with their back-testing.)

We'd love to hear the BLS's explanation for why they removed this disclaimer. Have the methods changed so radically (completely unannounced) that this flaw is now eliminated, or is this just more informational subterfuge from the government?

For the record, here is a pdf screenshot of today's version of the page, in case they do change it back.

A final question. Why remove the disclaimer if, as the latest employment benchmark revisions suggest, the BLS was only off by 14% in 2007?

Comments: Be the first to add a comment

add a comment | go to forum thread